Doubt with handwritten texts.

160 views
Skip to first unread message

Vagner Belfort

unread,
Jun 21, 2019, 1:56:25 AM6/21/19
to tesseract-ocr
Hello, I'm new to data science and I have a project to read the bank check, I can do any extraction of the typed characters, but I can not extract the handwritten texts.
My question, and is there any way to do this with tesserato?
I researched and found nothing about it. Thank you.
Attached has an example of what I'm trying to do.
01.jpeg

Timothy Snyder

unread,
Jun 21, 2019, 9:14:23 AM6/21/19
to tesser...@googlegroups.com
It's not possible out-of-the-box with Tesseract but I've reached ~90% accuracy so far on a handwriting model I'm working on. Check out projects like IAM, EMNIST, and UNIPEN to start collecting handwriting data/images.

You will probably want to segment the handwritten text off the check and apply a separate model trained with handwriting samples.

--
You received this message because you are subscribed to the Google Groups "tesseract-ocr" group.
To unsubscribe from this group and stop receiving emails from it, send an email to tesseract-oc...@googlegroups.com.
To post to this group, send email to tesser...@googlegroups.com.
Visit this group at https://groups.google.com/group/tesseract-ocr.
To view this discussion on the web visit https://groups.google.com/d/msgid/tesseract-ocr/bb853a71-5e33-4db4-bf13-dd876d7f81fd%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Vagner Belfort

unread,
Jun 21, 2019, 10:00:53 AM6/21/19
to tesseract-ocr
So I have to do the training of each check on an "IAM" data set? Thank you.

Em sexta-feira, 21 de junho de 2019 10:14:23 UTC-3, Timothy Snyder escreveu:
Não é possível usar o Tesseract com a Tesseract, mas alcancei até agora 90% de precisão em um modelo de manuscrito em que estou trabalhando. Confira projetos como IAM, EMNIST e UNIPEN para começar a coletar dados / imagens de manuscrito.

Você provavelmente desejará segmentar o texto manuscrito do cheque e aplicar um modelo separado treinado com amostras de manuscrito.

Sex, 21 de junho de 2019 às 01:56 Vagner Belfort < vagn ... @ gmail.com > escreveu:
Olá, sou novo em ciência de dados e tenho um projeto para ler o cheque bancário, posso fazer qualquer extração dos caracteres digitados, mas não consigo extrair os textos manuscritos.
Minha pergunta, e há alguma maneira de fazer isso com tesserato?
Eu pesquisei e não encontrei nada sobre isso. Obrigado.
Anexado tem um exemplo do que estou tentando fazer.

-
Você recebeu esta mensagem porque está inscrito no grupo "tesseract-ocr" dos Grupos do Google.
Para cancelar a inscrição deste grupo e deixar de receber emails, envie um email para tesser ... @ googlegroups.com .
Para enviar mensagens para este grupo, envie um e-mail para tesser ... @ googlegroups.com .
Visite este grupo em https://groups.google.com/ group / tesseract-ocr .
Para ver esta discussão na Web, visite https://groups.google.com/d/ msgid / tesseract-ocr / bb853a71-5e33-4db4-bf13-dd876d7f81fd% 40googlegroups.com .
Para obter mais opções, visite https://groups.google.com/d/ optout .

anonymous

unread,
Nov 1, 2019, 4:49:52 AM11/1/19
to tesseract-ocr
Could you please provide your source-code/GitHub link, so I can use it to recognize handwritten texts.

Brad Blood

unread,
Nov 4, 2019, 1:58:11 PM11/4/19
to tesseract-ocr
You might try this software. According to this, they specifically mention / illustrate OCR-ing handwriting on checks (look about halfway down the page): https://www.bisok.com/grooper-data-capture-method-features/multi-pass-ocr/

GN

unread,
Nov 7, 2019, 11:19:17 PM11/7/19
to tesseract-ocr
I would be also interested to check your github repo. Thanks!
Reply all
Reply to author
Forward
0 new messages